MIB–MIP is a mycoplasma system that captures and cleaves immunoglobulin G

Yonathan Arfia,b,1, Laetitia Minderc,d, Carmelo Di Primoe,f,g, Aline Le Royh,i,j, Christine Ebelh,i,j, Laurent Coquetk, Stephane Claveroll, Sanjay Vasheem, Joerg Joresn,o, Alain Blancharda,b, and Pascal Sirand-Pugneta,b

aINRA (Institut National de la Recherche Agronomique), UMR 1332 Biologie du Fruit et Pathologie, F-33882 Villenave d’Ornon, France; bUniversity of Bordeaux, UMR 1332 Biologie du Fruit et Pathologie, F-33882 Villenave d’Ornon, France; cInstitut Européen de Chimie et Biologie, UMS 3033, University of Bordeaux, 33607 Pessac, France; dInstitut Bergonié, SIRIC BRIO, 33076 Bordeaux, France; eINSERM U1212, ARN Regulation Naturelle et Artificielle, 33607 Pessac, France; fCNRS UMR 5320, ARN Regulation Naturelle et Artificielle, 33607 Pessac, France; gInstitut Européen de Chimie et Biologie, University of Bordeaux, 33607 Pessac, France; hInstitut de Biologie Structurale, University of Grenoble Alpes, F-38044 Grenoble, France; iCNRS, Institut de Biologie Structurale, F-38044 Grenoble, France; jCEA, Institut de Biologie Structurale, F-38044 Grenoble, France; kCNRS UMR 6270, Plateforme PISSARO, Institute for Research and Innovation in Biomedicine - Normandie Rouen, Normandie Université, F-76821 Mont-Saint-Aignan, France; lProteome Platform, Functional Genomic Center of Bordeaux, University of Bordeaux, F-33076 Bordeaux Cedex, France; mJ. Craig Venter Institute, Rockville, MD 20850; nInternational Livestock Research Institute, 00100 Nairobi, Kenya; and oInstitute of Veterinary Bacteriology, University of Bern, CH-3001 Bern, Switzerland

Edited by Roy Curtiss III, University of Florida, Gainesville, FL, and approved March 30, 2016 (received for review January 12, 2016) Mycoplasmas are “minimal” bacteria able to infect humans, wildlife, introduced into naive herds (8). In goats and sheep, mycoplasma and a large number of economically important livestock species. My- infections are predominantly caused by Mycoplasma capricolum coplasma infections include a spectrum of clinical manifestations subsp. capricolum, Mycoplasma capricolum subsp. capripneumoniae, ranging from simple fever to fulminant inflammatory diseases with and Mycoplasma mycoides subsp. capri. These pathogens cause acute high mortality rates. These infections are mostly chronic, suggesting to peracute diseases, often lethal, including contagious caprine that mycoplasmas have developed means to evade the host immune pleuropneumonia (9). Mycoplasmas also infect swine (Mycoplasma response. Here we present and functionally characterize a two-pro- tein system from Mycoplasma mycoides subspecies capri that is in- hyopneumoniae, Mycoplasma hyorhinis) (10) and economically im- volved in the capture and cleavage of IgG. The first component, portant Galliformes, such as chicken and turkey (Mycoplasma galli- Mycoplasma Ig binding protein (MIB), is an 83-kDa protein that is able septicum, Mycoplasma synoviae,andMycoplasma iowae)(11). to tightly bind to the Fv region of a wide range of IgG. The second Overall, the genetic bases of mycoplasma pathogenicity remain component, Mycoplasma Ig protease (MIP), is a 97-kDa serine prote- unclear, but they appear to lack the effectors exported through ase that is able to cleave off the VH domain of IgG. We demonstrate specialized secretion systems found in numerous bacterial pathogens that MIB is necessary for the proteolytic activity of MIP. Cleavage of (12). In addition, the number of toxins or other toxic compounds IgG requires a sequential interaction of the different partners of the produced by mycoplasmas is limited. Among the known exceptions is system: first MIB captures the IgG, and then MIP is recruited to the the capacity of several mycoplasmas to produce hydrogen peroxide MIB–IgG complex, enabling protease activity. MIB and MIP are (H2O2) as a by-product of glycerol metabolism (13, 14). For a small encoded by two genes organized in tandem, with homologs number of Mycoplasma species, the secretion of virulence-mediating found in the majority of pathogenic mycoplasmas and often in My- multiple copies. Phylogenetic studies suggest that genes encod- proteins was demonstrated. For instance, the rodent pathogen ing the MIB–MIP system are specific to mycoplasmas and have coplasma arthritidis secretes the effector Mycoplasma arthritidis mi- been disseminated by horizontal gene transfer. These results togen, an immunomodulatory protein that demonstrates classical highlight an original and complex system targeting the host im- superantigenic properties, resulting in an excessive activation of munoglobulins, playing a potentially key role in the immunity the immune system (15). Another example is the production by evasion by mycoplasmas. Significance mycoplasmas | immunoglobulin | protease Mycoplasmas are minimal pathogenic bacteria able to infect ycoplasmas are small-sized bacteria belonging to the class humans and a wide range of economically important animals; MMollicutes (1). These organisms are characterized by a fast as such, they are major causes of concern in the medical and evolution that has been marked by a drastic genome reduction, veterinary fields. These pathogens often lead to chronic in- leading to current mycoplasmas having a reduced coding capacity, fections, and their mechanisms of immunity evasion are poorly no cell wall, and a limited number of metabolic pathways (2, 3). As a known. Here we describe a two-protein system from the ru- result, they are obligate parasites and are found in vertebrate animals minant pathogen Mycoplasma mycoides subspecies capri that including mammalian, avian, reptilian, and piscine hosts (4). is involved in the capture and cleavage of antibodies. MIB is Despite their apparent simplicity, many mycoplasmas are path- able to capture the antibodies and to subsequently recruit MIP, ogenic for humans and a wide range of animals, and are major a protease that is able to cleave the antibody heavy chain. The – causes of concern in both the medical and veterinary fields (5). In MIB MIP system appears to be widespread among pathogenic humans, two of the most significant pathogens are Mycoplasma mycoplasmas and is potentially a key player for the virulence pneumoniae and Mycoplasma genitalium, infecting the respiratory and immunity evasion mechanisms of these bacteria. and urogenital tracts, respectively; they are mostly associated with Author contributions: Y.A., S.V., J.J., A.B., and P.S.-P. designed research; Y.A., L.M., C.D.P., diseases characterized by high morbidity and low mortality, like A.L.R., C.E., L.C., and S.C. performed research; Y.A., L.M., C.D.P., A.L.R., C.E., L.C., and S.C. interstitial pneumonia or urethritis (6, 7). In farm animals, myco- analyzed data; and Y.A. and P.S.-P. wrote the paper. plasmas have a high economic impact, thus warranting intensive The authors declare no conflict of interest. veterinary surveillance and regulation. In cattle, one of the diseases This article is a PNAS Direct Submission. listed by the World Organization for Animal Health is the conta- Freely available online through the PNAS open access option. gious bovine pleuropneumonia (CBPP) caused by Mycoplasma 1To whom correspondence should be addressed. Email: [email protected]. mycoides subspecies (subsp.) mycoides. CBPP is still highly preva- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. lent in Africa, and its mortality rate can reach up to 50% when 1073/pnas.1600546113/-/DCSupplemental.

5406–5411 | PNAS | May 10, 2016 | vol. 113 | no. 19 www.pnas.org/cgi/doi/10.1073/pnas.1600546113 Downloaded by guest on September 25, 2021 M. pneumoniae of the community acquired respiratory distress syn- drome toxin, an ADP ribosylating toxin with vacuolating activity (16). Though some mycoplasma infections can lead to acute in- flammatory episodes, most are chronic, suggesting that these bacteria have developed strategies to face and to evade the sophisticated immune system of mammals (17). The most widespread mechanism involves the antigenic variation of surface proteins, including immu- nodominant lipoproteins. In many cases, the antigens that are subject to variation are associated with cytoadherence (18, 19). Indirect evi- dences also point toward the existence of a molecular mimicry mechanism in M. pneumoniae, because infections by this mycoplasma have been followed in some cases by acute autoimmune diseases (20). It has been suggested that some mycoplasmas can directly target Fig. 1. Genome context of protein M homologs coding sequences. (Upper) the host’s immunoglobulins as a means to evade the immune sys- Schematic diagram of the genomic regions surrounding the genes MG281 in tem. In Ureaplasma, the cleavage of the human IgA1 at the hinge M. genitalium (blue) and MMCAP2_0583 (red) and MMCAP2_0582 (dark region, releasing intact Fab and Fc fragments, has been reported. It green) in M. mycoides subsp. capri. Genes homologous to MMCAP2_0583 are was also shown that isolates of Ureaplasma could cleave the IgA1 of colored in light red, and genes homologous to MMCAP2_0582 are colored in their host species, but not IgA1 of other species (21–23). This ac- light green. (Lower) Schematic representation of the domain architecture of tivity appears to be linked to a serine protease, but the source of this protein M and MMCAP_0583. activity remains cryptic, because no gene candidate could be iden- tified. In the poultry pathogens M. synoviae and M. gallisepticum,a this cluster are expressed in Mmc during growth on serum-based similar activity was attributed to the cysteine protease CysP (24). media (Fig. S1). To functionally characterize these two protein More recently, a new type of antibody-binding protein found in groups, we focused our study on two genes: MMCAP2_0582 and M. genitalium was described. This protein, named protein M, is extra- MMCAP2_0583. Analysis of MMCAP2_0583 using HHpred (27) cellular, membrane-anchored, and exhibits an antibody-binding to find remote homologs predicted a very strong structural simi- mechanism distinct from other known antibody-binding proteins larity with the functional domain MTD of protein M from like protein A from Staphylococcus aureus and protein G from M. genitalium despite very limited sequence similarities (Fig. S2). streptococcal bacteria (25). Protein M binds with a high affinity to This domain homologous to protein M is located on the C-terminal multiple types of human and nonhuman IgG, through attachment side of MMCAP2_0583, whereas the N-terminal side corresponded to conserved portions of the V domain of k and λ light chains. As L to a 250-aa domain with no homolog (Fig. 1). a result, protein M has the ability to prevent the antibody–antigen To study the function of MMCAP2_0583, the recombinant union, and could be part of an immunity evasion system based on antibody neutralization. Clear homologs to protein M were only protein r-0583 was produced by replacing the transmembrane identified in the closely related species M. pneumoniae,suggesting domain with a histidine tag. The protein r-0583 was expressed in that the system is restricted to a few species. Escherichia coli and purified to homogeneity. An initial ELISA In this study, we have identified a gene of M. mycoides subsp. experiment demonstrated that r-0583 is able to interact with a capri GM12 encoding a putative Mycoplasma Ig binding protein component of naive goat serum (Fig. S3). To identify this serum (MIB) presenting an internal domainwithastronglypredicted component, a pull-down experiment was performed using r-0583 structural similarity to protein M despite very low sequence con- as bait and naive goat serum as prey (Fig. 2A). Analysis of the servation. Consistently, we show that MIB acts as a high-affinity IgG eluted fractions showed that two proteins were purified from the binding protein. In addition, we have identified a gene encoding a serum through binding to r-0583. These proteins were identified – putative Mycoplasma Ig protease (MIP) immediately downstream to by liquid chromatography tandem mass spectrometry (LC-MS/ the MIB encoding gene. We demonstrate that MIP specifically MS) as the heavy and light chain of IgG, in accordance with their cleaves the IgG heavy chain at an unconventional site, located be- apparent molecular weights of 55 kDa and 25 kDa, respectively. The same experiment was performed using sera from various tween the VH and CH3 domains. Our data also reveal that prior binding of MIB to the IgG substrate is required for MIP activity. mammals, and showed that r-0583 is also able to interact with The MIB–MIP system is encoded by a pair of genes often found in IgG from sheep, bovine, horse, mouse, and human sources (Fig. several copies in a wide range of mycoplasmas that infect various S3). Interestingly, r-0583 is not able to bind to immunoglobulins hosts. Our bioinformatics analysis predicts horizontal gene transfer from rabbit serum. This result was confirmed by an ELISA of the MIB–MIP system between mycoplasmas sharing the same comparing the interaction between r-0583 and purified IgG from host, suggesting that MIB–MIP is a shared system involved in a goat or rabbit (Fig. S3). Overall, these results indicate that r-0583 global strategy to escape the host immune system. The identification shares the function of protein M and acts as a broad spectrum IgG of the MIB–MIP system opens a new avenue of research to better binding protein, hereafter termed MIB. understand the strategies used by minimal bacteria to evade the sophisticated immune system of complex hosts. MMCAP2_0582 Encodes an IgG Protease. The gene in tandem with the MIB-encoding MMCAP2_0583 was annotated as a putative Results serine protease, bearing the predicted domain of unknown func- MMCAP2_0583 Encodes an IgG Binding Protein. M. mycoides subsp. tion 31 (DUF31). Because the MMCAP2_0582 gene is genetically capri (Mmc) is a pathogenic mycoplasma associated with high linked to an IgG binding protein, we hypothesized that it could act mortality outbreaks of respiratory mycoplasmosis in small rumi- as an IgG protease. To validate this hypothesis, the recombinant nants (26). We have identified in Mmc an atypically organized protein r-0582 was produced by replacing the N-terminal trans- cluster of genes, comprised of four tandems (Fig. 1). The tandems membrane domain with a histidine tag. The protein r-0582 was all encode homologs of two putative proteins: a 97-kDa predicted expressed in E. coli, and subsequently purified to homogeneity. To serine protease (MMCAP2_0582, 0584, 0586, and 0588) and an test its activity, r-0582 was added to the fractions eluted from the 83-kDa protein of unknown function (MMCAP2_0583, 0585, pull-down experiment described above. After incubation, the in- 0587, and 0589). These proteins bear a transmembrane domain at tegrity of the IgG chains was checked by electrophoresis (Fig. 2B).

their N-terminal extremity and are predicted to be located at the The results showed that the addition of r-0582 lead to a cleavage MICROBIOLOGY cell surface. A RT-PCR experiment showed that all of the genes of of the heavy chain of the IgG, producing a fragment of ∼44 kDa.

Arfi et al. PNAS | May 10, 2016 | vol. 113 | no. 19 | 5407 Downloaded by guest on September 25, 2021 using IgG fragments F(ab′)2 and Fab instead of whole IgG (Fig. 3B), because these fragments bore only one of the possible cleavage sites. The results of this experiment showed that MIP is able to cleave both F(ab′)2 and Fab, confirming that the cleavage site is located between the VH and CH3 domains. The exact cleavage site was determined by sequencing the N-terminal amino acids of the 44-kDa fragment. The sequence obtained was VSSAST, and was identified at the N-terminal extremity of the CH3 domain (Fig. 3C). Interestingly, the triplet VSS is also present on the IgG of sheep, bovine, horse, human, and mouse (Fig. S6), which can all be cleaved by MIP.

The MIB–MIP System Requires a Sequential Assembly. To better un- derstand the interplay between MIB and MIP, and in particular to comprehend the inability of the protease to cleave the IgG alone, we studied the protein–protein interactions between the different Fig. 2. Functional analysis of r-0583 and r-0582. (A) Pull-down interaction assay partners. Using surface plasmon resonance (SPR) in a qualitative between r-0583 and naïve goat serum. r-0583 was bound to a resin and in- approach (Fig. 4A), we showed that r-0582 is unable to interact with cubated with serum. After extensive washing, bound proteins were eluted and analyzed by SDS/PAGE in reducing (Left) and nonreducing conditions (Right). either r-0583 or with IgG. Interestingly, when an r-0583/IgG com- Control experiments included the bait protein r-0583 without serum (lane: plex is preassembled, r-0582 is able to bind very tightly to this r-0583) and the naked resin incubated with the serum (lane: serum). Black arrow, compound. The kinetic parameters for this interaction were de- IgG chains (55 kDa, heavy; 25 kDa, light; ∼250 kDa, whole IgG). (B) Effect of termined using the inactive variant r-0582A (Fig. S7). We elected to r-0582 on IgG pulled down from goat serum. Proteins eluted from the pull-down use the inactive form of MIP to remove from the system the loss of experiments were mixed with purified r-0582 (lane: pull down + r-0582) and mass potentially caused by the cleavage of the VH domain, which incubated for 15 min at 30 °C. Samples were checked by SDS/PAGE in reducing could interfere with the SPR signal. This experiment determined (Left) and nonreducing conditions (Right). Black star, cleaved IgG heavy chain. 5 −1 −1 that KD = 0.37 nM, with ka = 3.19 × 10 s ·M and kd = 1.18 × (C) Cleavage of purified goat IgG by r-0582. Samples were checked by SDS/PAGE − − 10 4·s 1, indicative of a very strong interaction between MIP and and the IgG fragments were detected by Western blot. Nonspecific detection of – r-0583 was caused by its nonimmunological binding of the primary antibody. the MIB IgG complex. To determine the stoichiometry of the in- teraction, we measured the absolute molar mass of the MIB– IgG–MIP complex using size-exclusion chromatography coupled This result was confirmed by performing the cleavage experiment to multi-angle light scattering (SEC-MALS). After evaluating the with purified goat IgG, and detecting the Ig fragments with a specific mass of the individual components and confirming that their basic antibody (Fig. 2C). Interestingly, this experiment also showed that states are monomeric (Fig. S8), we analyzed the complexes IgG/ combining r-0582 with the IgG alone is not sufficient for cleavage to r-0583 and IgG/r-0583/r-0582 (Fig. 4B). The absolute molar mass occur, and that the presence MIB is necessary for the protease ac- of the IgG/r-0583 complex was determined at 303 kDa, which is tivity. The proteolytic cleavage of pulled-down IgG was also per- consistent with a 1 × IgG – 2 × r-0583 stoichiometry [expected formed with serum samples from various mammals (Fig. S4). This experiment showed that, in the presence of MIB, r-0582 is also able to cleave the IgG heavy chain from sheep, bovine, horse, mouse, and human, resulting in a 44-kDa fragment. Overall, these results point toward an IgG protease function for r-0582, hereafter termed MIP.

MIP Is a Serine Protease. Predictions based on sequence similarity suggested that MIP is a serine protease. To validate this hypothesis, we performed the IgG cleavage assay described above using MIP preincubated with PMSF, a serine protease inhibitor that reacts specifically with the active catalytic serine (Fig. S4). Unexpectedly, MIP was still able to cleave IgG in the presence of the inhibitor. The same result was obtained when using a commercial mixture of protease inhibitors targeting the majority of protease types. None- theless, the serine protease annotation for MMCAP2_0582 appeared to be confirmed by an HHpred analysis, because a small C-terminal domain of the protein had multiple hits in serine pro- teases (Fig. S5). Interestingly, the catalytic serines of the hits were systematically aligned with the Serine727 of MIP. Furthermore, analysis of the homologs of MIP in multiple mycoplasma species showed that Serine727 and its immediate surroundings were highly 727 conserved (Fig. S5). Consequently, a Ser Ala mutant (r-0582A) Fig. 3. Identification of r-0582 proteolytic site. (A) Schematic representation of ′ was produced, expressed in E. coli and purified. This mutant was the domain architecture of IgG, F(ab )2 and Fab. CH, conserved domain heavy used in an IgG cleavage assay and showed no proteolytic activity chain (beige); VH, variable domain heavy chain (orange); CL, conserved domain (Fig. 2C), confirming that MIP is indeed a serine protease. light chain (light blue); VL, variable domain light chain (dark blue). Arrows, possible cleavage sites. (B) Proteolytic cleavage of IgG fragments by r-0582. The IgG Cleavage Site by MIP Is Located Between the V and C 3 Samples were checked by SDS/PAGE and the IgG fragments were detected by H H Western blot. Black arrow, heavy and light chains of immunoglobulins; black Domains. Digestion of IgG by MIP led to a 44-kDa fragment, star, cleaved IgG heavy chain; circled star, cleaved F(ab′)2 fragment; white star, resulting from the cleavage of the heavy chain. Based on the size cleaved Fab fragment. (C) Cleavage site of r-0582. The N-terminal sequence of of this polypeptide, two possible cleavage sites can be deduced: the the large cleaved fragment of IgG heavy chain (black star) was determined by first located between the CH1andCH2 domains, and the second Edman degradation (bold and underlined). This sequence is located between between VH and CH3(Fig.3A). We performed a cleavage assay the VH domain (light blue) and CH3 domain (beige) of goat IgG.

5408 | www.pnas.org/cgi/doi/10.1073/pnas.1600546113 Arfi et al. Downloaded by guest on September 25, 2021 Mollicutes class. DUF31-annotated genes are widespread among animal and human pathogenic mycoplasmas, and interestingly absent from plant pathogenic species. It is noteworthy that, in some species like M. hemocanis or M. genitalium,DUF31- annotated genes are found in multiple copies, whereas the MIB– MIP system is absent. The exact function of these DUF31 genes is still elusive, although amino acid conservation suggests a serine–protease function. Discussion In this study, we identified and functionally characterized a two- protein system from the small-ruminant pathogen Mycoplasma mycoides subsp. capri, involved in the capture and cleavage of IgG. The first component of the system is the protein MIB, which is able to bind tightly to IgG. Given that MIB has a very strong predicted structural similarity with protein M from M. genitalium,anda similar function, we assumed that its binding mechanism was simi- lar. Protein M binds to immunoglobulins predominantly through attachment to the conserved portions of the VL domain of the light chain (25). This model is consistent with the observation that two MIB molecules can bind to a single IgG. Furthermore, MIB was able to bind to the Fab and F(ab′)2 fragments, confirming its in- teraction with the FV regions of the immunoglobulins. The second element of the system is the serine protease MIP, which is able to cleave the heavy chain of IgG. Our data showed Fig. 4. Assembly of the IgG–MIB–MIP complex. (A) Interactions among MIB, that MIP is unable to interact with either isolated MIB or IgG, but MIP, and goat IgG. SPR was used to analyze the interactions among IgG, r-0583, can bind very strongly to the complex they form. The exact assembly and r-0582. In a first experiment (Left), r-0583 was immobilized on the sensor mechanism of the MIB–IgG-MIP complex is not yet resolved, and it chipandprobedwithr-0582(1μM), goat IgG (0.5 μM), and again with r-0582 is unclear whether MIP binds directly to a domain of MIB, to a (1 μM). In a second experiment (Right), goat IgG was immobilized on the sensor domain of IgG, or to both. Interestingly, we showed that MIP alone μ μ μ chipandprobedwithr-0582(1 M), r-0583 (1 M), and again with r-0582 (1 M). is not able to cleave the IgG heavy chain, suggesting that it is in- (B) Molecular weight of the IgG–MIB and IgG–MIB–MIP complexes. Absolute molecular masses of the IgG/r-0583 complex (blue line) and IgG/r-0582/r-0583 active by itself. MIP is insensitive to PMSF, despite the fact that this complex (black line) were determined by SEC-MALS. The horizontal lines corre- inhibitor inactivates virtually all serine protease and is a diagnostic spond to the MALS calculated masses. Average mass values were calculated over for this protease class. PMSF inhibits serine proteases by reacting the half-height width of the elution peaks. specifically with the catalytic serine and thus blocking the active site (28). We therefore hypothesized that the catalytic residue Ser727 is either not exposed to the solvent or not in the activated state before molar mass ∼140 + (2 × 83) = 306 kDa]. For the IgG/r-0583/r-0582 the interaction of MIP with its cognate partners. Upon binding to complex, the absolute molar mass was determined at 495 kDa, which the MIB–IgG complex it is probable that significant conformational is consistent with a 1 × IgG – 2 × r-0583 − 2 × r0582 stoichiometry changes occur in MIP, leading to the activation of the catalytic [expected molar mass ∼140 + (2 × 83) + (2 × 97) = 500 kDa]. serine or to its positioning in the vicinity of the IgG cleavage site. Overall, these results indicate that one MIB is able to bind to Further studies on the 3D structure of the different partners within each of the two Fab domains of IgG. The resulting structure is and outside of the complex should provide us with a better un- able to recruit two MIP, one for each MIB–Fab pair (Fig. 5). The derstanding of this intricate mechanism. The data we gathered also resulting complex appears to be stable over time. show that the protease is bound very tightly to its partners, and is not released after the cleavage occurs. This observation could be The MIB–MIP System Is Widespread in Mollicutes. After the identi- linkedtotheinvitroexperimentalset-upusedinthisstudy.The fication of MIB and MIP, we searched for homologous genes in MIB–MIP system might be missing one or more partners involved the genomes of 39 Mollicutes representative of the diversity of this in further processing of the antibody before its release. Potential class, both in terms of taxonomy and host tropism (Fig. 6). Our candidates include other proteases such as the three homologs of data showed that, in contrast to protein M, which is only found in MMCAP2_0582, the three other DUF31 annotated genes, or the a small number of pathogens, MIB and MIP homologs are found gene MMCAP2_0490, which encodes a distant homolog of the IgG in the majority of animal pathogenic mycoplasmas. The tandem architecture was found in most of the genomes containing the MIB–MIP system, and was often complemented by isolated MIB or MIP homologous genes at remote loci. The system was no- ticeably absent from plant pathogens such as Mesoplasma florum and Spiroplasma citri and the acholeplasma/phytoplasma group. It is also interesting to note that the MIB–MIP system and protein M appear to be mutually exclusive because they are not found in the same genomes. The only exception is the presence of a remote homolog of protein M in M. gallisepticum alongside clear homo- logs of the MIB–MIP system. Furthermore, we scanned the dif- ferent genomes for the presence of genes annotated as DUF31 that are not homologs of MIP-encoding genes. This Pfam domain Fig. 5. Model of the assembly of the MIB–IgG–MIP complex. MIB (red) is able to

was the only annotation initially available for MMCAP2_0582, and bind tightly to the FV region of the IgG (gray). Upon formation of the IgG–MIB

all of the MIP predicted homologs shared this domain annotation. complex, MIP (green) is recruited. Upon binding, the catalytic serine of MIP is MICROBIOLOGY This domain appears to be found almost exclusively in the exposed (purple) and the cleavage of the heavy chain occurs (blue lightning bolt).

Arfi et al. PNAS | May 10, 2016 | vol. 113 | no. 19 | 5409 Downloaded by guest on September 25, 2021 63.5–70% identity, with most of the sequence differences located in the N-terminal domain of unknown function. For MIP paralogs, the dissimilarities were even more important (75.8–79.1% simi- larity and 58.4–64% identity) but spread throughout the protein. The specific roles and interplay of the different homologs and paralogs of MIB and MIP, as well as their regulation, remains to be explored. Nonetheless, data gathered from two recent publi- cations provided initial evidence that the MIB–MIP system is expressed in vivo. The proteome of M. mycoides subsp. mycoides (Mmm), a species closely related to Mmc, was analyzed in vitro and in vivo. These studies showed that five of six homologs of MIB and MIP could be detected in Mmm (37) and confirmed that these proteins were located at the surface of the bacteria (38). Based on sequence similarity values, the MIP and MIB homologs expressed in Mmm correspond to the proteins 0582, 0583, 0584, 0587, and 0588 from Mmc. MIB–MIP system homologs were found in the majority of animal pathogenic Mollicutes genomes, including the human pathogens M. hominis, Ureaplasma parvum,andUreaplasma urealyticum (Fig. 6). By contrast, no MIB–MIP system could be predicted in any plant pathogenic phytoplasmas and spiroplasmas. The system was also missing in several mycoplasmas including the two human pathogenic species M. genitalium and M. pneumoniae. Interestingly, these two organisms possess the protein M that is a structurally related to MIB. Despite a clear structural relatedness, the amino acid sequences of MIB homologs and protein M are only very weakly similar, which suggests a fast and specific evolution of Fig. 6. Distribution of the MIB and MIP homologs in Mollicutes. The occurrence the protein M encoding gene in the few species where it is of a gene homolog of MIB, MIP, protein M, or a gene annotated DUF31 is in- found. Moreover, no protein homologous to MIP could be dicated by rectangles colored in red, green, dark blue, and light gray, respectively. identified for these two species, but multiple genes encoding Adjoining rectangles denote genes located on the same loci. Vertical alignment of proteins with a predicted DUF31 domain were predicted, which the symbols does not infer synteny. The host specificity of the Mollicutes species is could potentially be functional homologs of MIP—namely, IgG indicated as follows: black, animal host; cyan, human host; green, plant host. The proteases working in pair with protein M. phylogenetic tree was inferred using the maximum-likelihood method from the – concatenated multiple alignments of 79 proteins encoded by genes present at The wide distribution of the MIB MIP system in mycoplasmas one copy in each genome. Main phylogenetic groups are indicated: S, spi- infecting different animals and belonging to different phylogenetic roplasma; H, hominis; P, pneumoniae; and AAP, acholeplasma/phytoplasma. groups suggested a biological importance of this system regarding host–pathogen interactions. Phylogenetic analyses of the MIB and MIP (Fig. S9) proteins strongly suggested that horizontal gene protease CysP (24). Combined activities of these proteases might transfers (HGTs) occurred several times during the Mollicutes’ further degrade immunoglobulins and lead to a final release of the evolution. In particular, HGT were predicted among ruminant my- MIB–MIP components. coplasmas between the ancestor of the M. agalactiae/M. bovis group Another idiosyncrasy of the MIP protease is the location of and the ancestor of the mycoides cluster M. yeatsii/M. putrefaciens its cleavage site on the IgG. All of the bacterial Ig proteases group. Other HGTs of potential virulence factors have already been characterized to date have a cleavage site located in the hinge described among these groups (39). In urogenital human myco- region (29, 30), leading to the release of the Fc and Fab frag- plasmas, HGTs were predicted between M. hominis and the ancestor ments. It is surmised that the role of these proteases is to protect of the U. parvum/U. urealyticum group, in accordance with previous the bacteria from the host immune system by either preventing study (40). These genetic exchanges of the MIB–MIP system among the binding of the complement to Fc or masking the cell with a mycoplasmas colonizing the same host are in complete agreement cloak of host-like components (31). In the case of MIP, it is with the potential significance of this system in the evolutionary not clear whether the VH domain would be released from the process leading to the adaptation to a new host. – antibody after cleavage or not. The VH domain is bound non- Finally, we noted that the MIB MIP system homologs were covalently to the VL domain (32), but with relatively low disso- found exclusively in Mollicutes. It is interesting to find that − − ciation constants (10 5 − 10 8 M) that are not sufficient to keep such an intricate mechanism evolved in so-called “minimal” the domains associated at low concentrations of protein or bacteria, which lack some of the most basic cell functions. Al- under mildly destabilizing conditions (33). Loss of the VH do- though further studies are necessary to completely elucidate main would be highly detrimental for the function of the anti- this system and its in vivo implications, the identification of body as this domain bears three of the six complementarity MIB and MIP opens a new avenue of research to better un- determining regions (CDRs) involved in antigen recognition, derstand the pathogenicity of mycoplasmas and their immunity including CDR-H3 which is the key player in the antigen af- evasion mechanisms. finity (34–36). Therefore, the role of the MIB–MIP system Materials and Methods could be to render the IgG nonfunctional, unable to target their antigens, thus limiting the efficiency of the immune response. Cloning. MMCAP2_0582 and MMCAP2_0583 coding sequences were codon optimized for expression in E. coli. Synthetic genes were ordered from In the genome of Mmc, four pairs of genes encoding MIB and Eurofins Genomics and cloned in the pET28a(+) vector (Novagen). Details for MIP proteins were found in tandem (Fig. 1). The intraspecies this and subsequent methods can be found in SI Materials and Methods. diversity between these multiple variants appeared to be quite high (Tables S1–S4). Amino acids conservation between MIB Recombinant Proteins Expression and Purification. Recombinant protein were and MIB paralogs was limited to 78.4–83.5% similarity and expressed in E. coli BL21(DE3) (Novagen) grown in autoinduction medium at

5410 | www.pnas.org/cgi/doi/10.1073/pnas.1600546113 Arfi et al. Downloaded by guest on September 25, 2021 20 °C. Protein purification was performed on a Nickel-NTA column (GE Health- Protein–Protein Interaction Analysis. Analysis of the interactions was per- care Life Sciences) followed by polishing and buffer exchange by gel filtration on formed by SPR on a Biacore T200 instrument using CM5 S Series sensor chips a Superdex 200 column (GE Healthcare Life Sciences). (GE Healthcare Life Sciences). Ligands were coupled to the chip surface using the Biacore amine coupling protocol, and analytes were injected at a flow of Pull-Down Assays. Animal sera were obtained either from commercial vendors 20 μL·min−1. All of the analyses were performed at 30 °C, in PBS pH 7.4 (Sigma) or as gifts from academic research laboratories. Nickel-NTA agarose beads Tween 20 0.05% buffer. Kinetic parameters were determined by performing (GE Healthcare Life Sciences) coated with purified r-0583 were incubated with a single-cycle kinetic experiment, and data were fitted to a 1:1 Langmuir model. serum for 2 h at room temperature. Unbound proteins were thoroughly washed with buffer containing 20 mM imidazole, and bound proteins were subsequently SEC-MALS. Absolute molecular masses were determined using SEC coupled to eluted using buffer containing 250 mM imidazole. a MALS detector. SEC was performed using a HPLCy system and a Superose − 6 10/300 GL column (GE) thermostated at 30 °C, at a flow rate of 0.5 mL·min 1. MS Identification. Protein identity was determined using a LC-MS/MS analysis The elution profiles were followed by refractive index, and by static light pipeline. Proteins were first reduced then alkylated before undergoing tryptic digestion overnight at 37 °C. Extracted peptides were analyzed on an Ultimate scattering using an Optilab rEX, a miniDAWN TREOS, respectively (Wyatt 3000 nanoLC system (Dionex) coupled to a nanospray LTQ-Orbitrap XL mass Technology). Molar masses are given as mean over the half-height width of spectrometer (Thermo Finnigan). Data were searched by SEQUEST through Pro- the elution peaks. teome Discoverer 1.4 (Thermo Fisher Scientific Inc.) against databases containing amino acid sequences of Ig extracted from the abYsis database (www.bioinf.org. ACKNOWLEDGMENTS. The authors thank Sybille Duret and Dr. Jan uk/abysis). Naessens for their assistance, and Dr. Maude Ele-Deverttus for the discussions. Goat sera were kindly provided by Dr. Tardy. We thank the structural biophysico-chemistry facility (UMS 3033/US001) of the Institut Ig Cleavage Assays. Purified goat IgG was purchased from Sigma, and purified Européen de Chimie et Biologie (Pessac, France) for access to the Biacore ′ goat F(ab )2 and Fab were purchased from Jackson ImmunoResearch. All cleavage T200 instrument that was acquired with the support of the Conseil assays were performed at 30 °C by mixing all of the components in PBS buffer Régional d’Aquitaine, the Groupement d’Intérêt Scientifique–Infrastrutures pH 7.4. The protease was always the last element added to the reaction mixture. en Biologie Sante et Agronomie, and the Cellule Hôtels à Projets of the CNRS. The platform of the Grenoble Instruct Center (Integrated Structural Western Blots. Proteins separated by SDS/PAGE were transferred on a ni- Biology Grenoble; UMS 3518 CNRS-CEA-UJF-EMBL) is supported by the French Infrastructure for Integrated Structural Biology (ANR-10-INSB-05- trocellulose membrane using a TE-77 semidry transfer system (GE Healthcare 02) and Grenoble Alliance for Integrated Structural Cell Biology (ANR-10- Life Sciences). Detection was performed using a primary antibody (mouse LABX-49-01) within the Grenoble Partnership for Structural Biology. This + anti-goat IgG H L; Jackson ImmunoResearch) and a secondary antibody work was funded by the National Science Foundation Basic Research to (goat anti-mouse Fc HRP coupled) and the SuperSignal West Pico Chemilu- Enable Agricultural Development Program Grant 1110151. L.M. is sup- minescent Substrate (Thermo Scientific). ported by the SIRIC BRIO.

1. Brown DR, Whitcomb RF, Bradbury JM (2007) Revised minimal standards for de- 21. Spooner RK, Russell WC, Thirkell D (1992) Characterization of the immunoglobulin A scription of new species of the class Mollicutes (division Tenericutes). Int J Syst Evol protease of Ureaplasma urealyticum. Infect Immun 60(6):2544–2546. Microbiol 57(Pt 11):2703–2719. 22. Paralanov V, et al. (2012) Comparative genome analysis of 19 Ureaplasma urealyticum 2. May M, Balish MF, Blanchard A (2014) The Prokaryotes: Firmicutes and Tenericutes,eds and Ureaplasma parvum strains. BMC Microbiol 12:88. Rosenberg E, DeLong EF, Lory S, Stackebrandt E, Thompson F (Springer, Berlin), pp 515–550. 23. Robertson JA, Stemler ME, Stemke GW (1984) Immunoglobulin A protease activity of 3. Sirand-Pugnet P, Citti C, Barré A, Blanchard A (2007) Evolution of mollicutes: Down a Ureaplasma urealyticum. J Clin Microbiol 19(2):255–258. bumpy road with twists and turns. Res Microbiol 158(10):754–766. 24. Cizelj I, et al. (2011) Mycoplasma gallisepticum and Mycoplasma synoviae express a 4. Brown DR, et al. (2011) Genus I. mycoplasma. Bergey’s Manual of Systematic Bacte- cysteine protease CysP, which can cleave chicken IgG into Fab and Fc. Microbiology riology, ed Krieg NR, et al. (Springer, New York), 2nd Ed, Vol 4, pp 575–613. 157(Pt 2):362–372. 5. Citti C, Blanchard A (2013) Mycoplasmas and their host: Emerging and re-emerging 25. Grover RK, et al. (2014) A structurally distinct human mycoplasma protein that ge- – minimal pathogens. Trends Microbiol 21(4):196 203. nerically blocks antigen-antibody union. Science 343(6171):656–661. 6. Atkinson TP, Balish MF, Waites KB (2008) Epidemiology, clinical manifestations, 26. Hernandez L, et al. (2006) Mycoplasma mycoides subsp. capri associated with goat pathogenesis and laboratory detection of Mycoplasma pneumoniae infections. FEMS respiratory disease and high flock mortality. Can Vet J 47(4):366–369. – Microbiol Rev 32(6):956 973. 27. Remmert M, Biegert A, Hauser A, Söding J (2011) HHblits: Lightning-fast iterative 7. Taylor-Robinson D, Jensen JS (2011) Mycoplasma genitalium: From chrysalis to mul- protein sequence searching by HMM-HMM alignment. Nat Methods 9(2):173–175. – ticolored butterfly. Clin Microbiol Rev 24(3):498 514. 28. Hedstrom L (2002) Serine protease mechanism and specificity. Chem Rev 102(12):4501–4524. 8. Thiaucourt F, Aboubakar Y, Wesonga H, Manso-Silvan L, Blanchard A (2004) Conta- 29. Potempa J, Pike RN (2009) Corruption of innate immunity by bacterial proteases. gious bovine pleuropneumonia vaccines and control strategies: recent data. Dev Biol J Innate Immun 1(2):70–87. – (Basel) 119:99 111. 30. Mistry D, Stockley RA (2006) IgA1 protease. Int J Biochem Cell Biol 38(8):1244–1248. 9. Manso-Silván L, Dupuy V, Chu Y, Thiaucourt F (2011) Multi-locus sequence analysis of 31. Brezski RJ, Jordan RE (2010) Cleavage of IgGs by proteases associated with invasive Mycoplasma capricolum subsp. capripneumoniae for the molecular epidemiology of diseases: An evasion tactic against host immunity? MAbs 2(3):212–220. contagious caprine pleuropneumonia. Vet Res (Faisalabad) 42:86. 32. Horne C, Klein M, Polidoulis I, Dorrington KJ (1982) Noncovalent association of heavy 10. Simionatto S, Marchioro SB, Maes D, Dellagostin OA (2013) Mycoplasma hyopneu- and light chains of human immunoglobulins. III. Specific interactions between VH and moniae: From disease to vaccine development. Vet Microbiol 165(3-4):234–242. VL. J Immunol 129(2):660–664. 11. Stipkovits L, Kempf I (1996) Mycoplasmoses in poultry. Rev Sci Tech 15(4):1495–1525. 33. Röthlisberger D, Honegger A, Plückthun A (2005) Domain interactions in the Fab 12. Mattoo S, Lee YM, Dixon JE (2007) Interactions of bacterial effector proteins with host fragment: A comparative evaluation of the single-chain Fv and Fab format en- proteins. Curr Opin Immunol 19(4):392–401. gineered with variable domains of different stability. J Mol Biol 347(4):773–789. 13. Jenkins C, Samudrala R, Geary SJ, Djordjevic SP (2008) Structural and functional 34. Bujotzek A, et al. (2015) Prediction of VH-VL domain orientation for antibody variable characterization of an organic hydroperoxide resistance protein from Mycoplasma domain modeling. Proteins 83(4):681–695. gallisepticum. J Bacteriol 190(6):2206–2216. 35. Ohno S, Mori N, Matsunaga T (1985) Antigen-binding specificities of antibodies are 14. Schmidl SR, et al. (2011) A trigger enzyme in Mycoplasma pneumoniae: Impact of the – glycerophosphodiesterase GlpQ on virulence and gene expression. PLoS Pathog 7(9): primarily determined by seven residues of VH. Proc Natl Acad Sci USA 82(9):2945 2949. e1002263. 36. Shirai H, Kidera A, Nakamura H (1999) H3-rules: Identification of CDR-H3 structures in – 15. Luo W, et al. (2008) Association of Mycoplasma arthritidis mitogen with lethal toxicity antibodies. FEBS Lett 455(1-2):188 197. but not with arthritis in mice. Infect Immun 76(11):4989–4998. 37. Weldearegay YB, et al. (2016) Proteomic characterization of pleural effusion, a spe- 16. Becker A, et al. (2015) Structure of CARDS toxin, a unique ADP-ribosylating and vacuolating cific host niche of Mycoplasma mycoides subsp. mycoides from cattle with contagious – cytotoxin from Mycoplasma pneumoniae. Proc Natl Acad Sci USA 112(16):5165–5170. bovine pleuropneumonia (CBPP). J Proteomics 131:93 103. 17. Simmons WL, Dybvig K (2007) How some mycoplasmas evade host immune responses. 38. Krasteva I, et al. (2014) Characterization of the in vitro core surface proteome of Microbe Mag 2(11):537–543. Mycoplasma mycoides subsp. mycoides, the causative agent of contagious bovine 18. Rosengarten R, Wise KS (1990) Phenotypic switching in mycoplasmas: Phase variation pleuropneumonia. Vet Microbiol 168(1):116–123. of diverse surface lipoproteins. Science 247(4940):315–318. 39. Sirand-Pugnet P, et al. (2007) Being pathogenic, plastic, and sexual while living with a 19. Ron Y, Flitman-Tene R, Dybvig K, Yogev D (2002) Identification and characterization nearly minimal bacterial genome. PLoS Genet 3(5):e75. of a site-specific tyrosine recombinase within the variable loci of Mycoplasma bovis, 40. Pereyre S, et al. (2009) Life on arginine for Mycoplasma hominis: Clues from its Mycoplasma pulmonis and Mycoplasma agalactiae. Gene 292(1-2):205–211. minimal genome and comparison with other human urogenital mycoplasmas. PLoS 20. Jacobs E, Bartl A, Oberle K, Schiltz E (1995) Molecular mimicry by Mycoplasma Genet 5(10):e1000677. pneumoniae to evade the induction of adherence inhibiting antibodies. J Med 41. Palau W, Di Primo C (2013) Simulated single-cycle kinetics improves the design of MICROBIOLOGY Microbiol 43(6):422–429. surface plasmon resonance assays. Talanta 114:211–216.

Arfi et al. PNAS | May 10, 2016 | vol. 113 | no. 19 | 5411 Downloaded by guest on September 25, 2021